Active Panoramic Referring Expression Segmentation on APRS (test)
75.4SRPanoSeeker
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| PanoSeekerExploration Strategy=VLM-based Agent (Active Exploration), Venue=None, Memory=EgoSphere2026.07 | 75.4 | 4.8 | 57 | 55.8 | |
| GPT-5.2Exploration Strategy=VLM-based Agent (Active Exploration), Venue=None, Memory=Text Log2026.07 | 69.1 | 6.2 | 36 | 53.2 | |
| VisionReasonerExploration Strategy=Static (Direct Panorama Input), Venue=ICLR'26, Memory=None2026.07 | 66.2 | 1 | — | 47.7 | |
| Gemini-3-FlashExploration Strategy=VLM-based Agent (Active Exploration), Venue=None, Memory=Text Log2026.07 | 64.9 | 6.8 | 31 | 51.4 | |
| LISAExploration Strategy=Static (Direct Panorama Input), Venue=CVPR'24, Memory=None2026.07 | 64.1 | 1 | — | 44.5 | |
| Qwen3-VL-30B-A3B-ThinkingExploration Strategy=VLM-based Agent (Active Exploration), Venue=ArXiv'25, Memory=Text Log2026.07 | 62.9 | 8 | 29 | 51 | |
| SAM4MLLMExploration Strategy=Static (Direct Panorama Input), Venue=ECCV'24, Memory=None2026.07 | 62.3 | 1 | — | 46.2 | |
| CRISExploration Strategy=Static (Direct Panorama Input), Venue=CVPR'22, Memory=None2026.07 | 55.4 | 1 | — | 39.2 | |
| VisionReasonerExploration Strategy=Heuristic (Pre-defined Scanning, Max 9 steps), Venue=ICLR'26, Memory=None2026.07 | 49.5 | 4.9 | 26 | 39.9 | |
| VLTExploration Strategy=Static (Direct Panorama Input), Venue=ICCV'21, Memory=None2026.07 | 46.7 | 1 | — | 34.7 | |
| LISAExploration Strategy=Heuristic (Pre-defined Scanning, Max 9 steps), Venue=CVPR'24, Memory=None2026.07 | 42.3 | 5.2 | 22 | 34.1 | |
| SAM4MLLMExploration Strategy=Heuristic (Pre-defined Scanning, Max 9 steps), Venue=ECCV'24, Memory=None2026.07 | 40.5 | 5.3 | 20 | 32.5 | |
| CRISExploration Strategy=Heuristic (Pre-defined Scanning, Max 9 steps), Venue=CVPR'22, Memory=None2026.07 | 32.1 | 4.4 | 17 | 26.4 | |
| VLTExploration Strategy=Heuristic (Pre-defined Scanning, Max 9 steps), Venue=ICCV'21, Memory=None2026.07 | 28.4 | 4.6 | 14 | 22.1 |